Ignore:
Timestamp:
02/27/13 15:02:50 (14 months ago)
Author:
Ales Erjavec <ales.erjavec@…>
Branch:
default
Message:

Cleanup of 'Widget catalog' documentation.

Fixed rst text formating, replaced dead hardcoded reference links (now using
:ref:), etc.

File:
1 edited

Legend:

Unmodified
Added
Removed
  • docs/widgets/rst/visualize/linearprojection.rst

    r11050 r11359  
    66.. image:: ../icons/LinearProjection.png 
    77 
    8 Various linear projection methods with explorative data analysis and intelligent data visualization enhancements. 
     8Various linear projection methods with explorative data analysis and 
     9intelligent data visualization enhancements. 
    910 
    1011Signals 
     
    2930 
    3031 
    31 Warning: this widget combines a number of visualization methods that are currently in research. Eventually, it will break down to a set of simpler widgets, each implementing its own method. 
     32Warning: this widget combines a number of visualization methods that are 
     33currently in research. Eventually, it will break down to a set of simpler 
     34widgets, each implementing its own method. 
    3235 
    3336Description 
    3437----------- 
    3538 
    36 This widget provides an interface to a number of linear projection methods that all deal with class-labeled data and aim at finding the two-dimensional projection where instances of different classes are best separated. Consider, for a start, a projection of a <a href="">zoo.tab</a> data set (animal species and their features) shown below. Notice that it is breast-feeding (milk) and hair that nicely characterizes mamals from the other organisms, and that laying eggs is something that birds do. This specific visualization was obtained using FreeViz (`Demsar et al., 2007 <#Demsar2007>`_), while the widget also implements an interface to supervised principal component analysis (`Koren and Carmel, 2003 <#Koren2003>`_), partial least squares (for a nice introduction, see `Boulesteix and Strimmer, 2006 <Boulesteix2007>`_), and RadViz visualization and associated intelligent data visualization technique called VizRank (<a href=""></a>). 
     39This widget provides an interface to a number of linear projection methods that 
     40all deal with class-labeled data and aim at finding the two-dimensional 
     41projection where instances of different classes are best separated. Consider, 
     42for a start, a projection of a **zoo.tab** data set (animal species and their 
     43features) shown below. Notice that it is breast-feeding (milk) and hair that 
     44nicely characterizes mamals from the other organisms, and that laying eggs is 
     45something that birds do. This specific visualization was obtained using FreeViz 
     46([Demsar2007]_), while the widget also implements an interface to supervised 
     47principal component analysis ([Koren2003]_), partial least squares (for a nice 
     48introduction, see [Boulesteix2007]_), and RadViz visualization and 
     49associated intelligent data visualization technique called VizRank  
     50([Leban2006]_) 
    3751 
    3852.. image:: images/LinearProjection-Zoo.png 
     53   :alt: Lienar Projection on zoo data set 
    3954 
    40 Projection search methods are invoked from :obj:`Optimization Dialogs` in the :obj:`Main` tab. Other controls in this tab and controls in the :obj:`Settings` tab are just like those with other visualization widgets; please refer to a documentation of `Scatterplot <Scatterplot.html>`_ widget for further information. 
     55Projection search methods are invoked from :obj:`Optimization Dialogs` in the 
     56:obj:`Main` tab. Other controls in this tab and controls in the :obj:`Settings` 
     57tab are just like those with other visualization widgets; please refer to a 
     58documentation of :ref:`Scatter Plot` widget for further information. 
    4159 
    4260.. image:: images/LinearProjection-FreeViz.png 
    4361   :alt: FreeViz screen shot 
    4462 
    45 :obj:`FreeViz` button in :obj:`Main` tab opens a dialog from which four different methods are accessed. The first one is FreeViz, which uses a paradigm borrowed from particle physics: points in the same class attract each other, those from different class repel each other, and the resulting forces are exerted on the anchors of the attributes, that is, on unit vectors of each of the dimensional axis. The points cannot move (are projected in the projection space), but the attribute anchors can, so the optimization process is a hill-climbing optimization where at the end the anchors are placed such that forces are in equilibrium. The FreeViz optimization dialog is used to invoke the optimization process (:obj:`Optimize Separation`) or execute a single step of optimization (:obj:`Single Step`). The result of the optimization may depend on the initial placement of the anchors, which can be set in a circle, arbitrary or even manually (:obj:`Set anchor positions`). The later also works at any stage of optimization, and we recommend to play with this option in order to understand how a change of one anchor affects the positions of the data points. Controls in :obj:`Forces` box are used to set the parameters that define the type of the forces between the data points (see `Demsar et al., 2007 <#Demsar2007>`_). In any linear projection, projections of unit vector that are very short compared to the others indicate that their associated attribute is not very informative for particular classification task. Those vectors, that is, their corresponding anchors, may be hidden from the visualization using controls in :obj:`Show anchors` box. 
     63:obj:`FreeViz` button in :obj:`Main` tab opens a dialog from which four 
     64different methods are accessed. The first one is FreeViz, which uses a paradigm 
     65borrowed from particle physics: points in the same class attract each other, 
     66those from different class repel each other, and the resulting forces are 
     67exerted on the anchors of the attributes, that is, on unit vectors of each of 
     68the dimensional axis. The points cannot move (are projected in the projection 
     69space), but the attribute anchors can, so the optimization process is a 
     70hill-climbing optimization where at the end the anchors are placed such that 
     71forces are in equilibrium. The FreeViz optimization dialog is used to invoke 
     72the optimization process (:obj:`Optimize Separation`) or execute a single step 
     73of optimization (:obj:`Single Step`). The result of the optimization may depend 
     74on the initial placement of the anchors, which can be set in a circle, 
     75arbitrary or even manually (:obj:`Set anchor positions`). The later also works 
     76at any stage of optimization, and we recommend to play with this option in 
     77order to understand how a change of one anchor affects the positions of the 
     78data points. Controls in :obj:`Forces` box are used to set the parameters that 
     79define the type of the forces between the data points (see [Demsar2007]_). 
     80In any linear projection, projections of unit vector that are very short 
     81compared to the others indicate that their associated attribute is not very 
     82informative for particular classification task. Those vectors, that is, their 
     83corresponding anchors, may be hidden from the visualization using controls in 
     84:obj:`Show anchors` box. 
    4685 
    47 The other two, quite prominent visualization methods, are accessible through FreeViz's :obj:`Dimensionality Reduction` tab (not shown here). These includes supervised principal component analysis and partial least squares method. The general objection of these two approaches is the same as for FreeViz (find a projection that separates data instances of different class), but the results - because of different optimization methods and differences in their bias - may be quite different. 
     86The other two, quite prominent visualization methods, are accessible through 
     87FreeViz's :obj:`Dimensionality Reduction` tab (not shown here). These includes 
     88supervised principal component analysis and partial least squares method. 
     89The general objection of these two approaches is the same as for FreeViz 
     90(find a projection that separates data instances of different class), but the 
     91results - because of different optimization methods and differences in their 
     92bias - may be quite different. 
    4893 
    49 The fourth projection search technique that can be accessed from this widget is VizRank search algorithm with RadViz visualization (Leban et al. (2006)). This is essentially the same visualization and projection search method as implemented in `Radviz <Radviz>`_. 
     94The fourth projection search technique that can be accessed from this widget 
     95is VizRank search algorithm with RadViz visualization ([Leban2006]_). This is 
     96essentially the same visualization and projection search method as implemented 
     97in :ref:`Radviz`. 
    5098 
    51 Like other point-based visualization widget, Linear Projection also includes explorative analysis functions (selection of data instances and zooming). See documentation for :doc:`Scatterplot <scatterplot>` widget for documentation of these as implemented in :obj:`Zoom / Select` toolbox in the :obj:`Main` tab of the widget. 
     99Like other point-based visualization widget, Linear Projection also includes 
     100explorative analysis functions (selection of data instances and zooming). 
     101See documentation for :ref:`Scatter Plot` widget for documentation of these as 
     102implemented in :obj:`Zoom / Select` toolbox in the :obj:`Main` tab of the 
     103widget. 
    52104 
    53105 
     
    55107---------- 
    56108 
    57   - Demsar J, Leban G, Zupan B. FreeViz-An intelligent multivariate visualization approach to explorative analysis of biomedical data. J Biomed Inform 40(6):661-71, 2007. 
    58   - Koren Y, Carmel L. Visualization of labeled data using linear transformations, in: Proceedings of IEEE Information Visualization 2003 (InfoVis'03), 2003. `PDF <http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=3DDF0DB68D8AB9949820A19B0344C1F3?doi=10.1.1.13.8657&rep=rep1&type=pdf>`_ 
    59   - Boulesteix A-L, Strimmer K (2006) Partial least squares: a versatile tool for the analysis of high-dimensional genomic data, Briefings in Bioinformatics 8(1): 32-44. `Abstract <http://bib.oxfordjournals.org/cgi/content/abstract/8/1/32>`_ 
    60   - Leban, G., B. Zupan, et al. (2006). "VizRank: Data Visualization Guided by Machine Learning." Data Mining and Knowledge Discovery 13(2): 119-136. 
     109.. [Demsar2007] Demsar J, Leban G, Zupan B. FreeViz-An intelligent multivariate 
     110   visualization approach to explorative analysis of biomedical data. J Biomed 
     111   Inform 40(6):661-71, 2007. 
     112 
     113.. [Koren2003] Koren Y, Carmel L. Visualization of labeled data using linear 
     114   transformations, in: Proceedings of IEEE Information Visualization 2003 
     115   (InfoVis'03), 2003. `PDF <http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=3DDF0DB68D8AB9949820A19B0344C1F3?doi=10.1.1.13.8657&rep=rep1&type=pdf>`_ 
     116 
     117.. [Boulesteix2007] Boulesteix A-L, Strimmer K (2006) Partial least squares: 
     118   a versatile tool for the analysis of high-dimensional genomic data, 
     119   Briefings in Bioinformatics 8(1): 32-44.  
     120   `Abstract <http://bib.oxfordjournals.org/cgi/content/abstract/8/1/32>`_ 
     121 
     122.. [Leban2006] Leban, G., B. Zupan, et al. (2006). "VizRank: Data Visualization 
     123   Guided by Machine Learning." Data Mining and Knowledge Discovery 13(2): 
     124   119-136. 
Note: See TracChangeset for help on using the changeset viewer.